The proliferation of harmful content on online platforms is a major societal problem, which comes in many different forms, including hate speech, offensive language, bullying and harassment, misinformation, spam, violence, graphic content, sexual abuse, self-harm, and many others. Online platforms seek to moderate such content to limit societal harm, to comply with legislation, and to create a more inclusive environment for their users. Researchers have developed different methods for automatically detecting harmful content, often focusing on specific sub-problems or on narrow communities, as what is considered harmful often depends on the platform and on the context. We argue that there is currently a dichotomy between what types of harmful content online platforms seek to curb, and what research efforts there are to automatically detect such content. We thus survey existing methods as well as content moderation policies by online platforms in this light and suggest directions for future work.

Abusive language on online platforms is a major societal problem, often leading to important societal problems such as the marginalisation of underrepresented minorities. There are many different forms of abusive language such as hate speech, profanity, and cyber-bullying, and online platforms seek to moderate it in order to limit societal harm, to comply with legislation, and to create a more inclusive environment for their users. Within the field of Natural Language Processing, researchers have developed different methods for automatically detecting abusive language, often focusing on specific subproblems or on narrow communities, as what is considered abusive language very much differs by context. We argue that there is currently a dichotomy between what types of abusive language online platforms seek to curb, and what research efforts there are to automatically detect abusive language. We thus survey existing methods as well as content moderation policies by online platforms in this light, and we suggest directions for future work.

Use of artificial intelligence is growing and expanding into applications that impact people's lives. People trust their technology without really understanding it or its limitations. There is the potential for harm and we are already seeing examples of that in the world. AI researchers have an obligation to consider the impact of intelligent applications they work on. While the ethics of AI is not clear-cut, there are guidelines we can consider to minimize the harm we might introduce.

Advances in interface design using touch surfaces creates greater obstacles for blind and visually impaired users of technology. Conversational user interfaces offer a reasonable alternative for interactions and enable greater access and most importantly greater independence for the blind. This paper presents a case study of our work to develop a conversational user interface for accessibility for multifunction printers (MFP). It describes our approach to conversational interfaces in general and the specifics of the solution we created for MFPs. It also presents a user study we performed to assess the solution and guide our future efforts.

Modern trends in interface design for office equipment using controls on touch surfaces create greater obstacles for blind and visually impaired users and contribute to an environment of dependency in work settings. We believe that \textit{conversational user interfaces} (CUIs) offer a reasonable alternative to touchscreen interactions enabling more access and most importantly greater independence for blind knowledge workers. We present a case study of our work to develop a conversational user interface for accessibility for multifunction printers. We also describe our approach to conversational interfaces in general, which emphasizes task-based collaborative interactions between people and intelligent agents, and we detail the specifics of the solution we created for multifunction printers. To guide our design, we worked with a group of blind and visually impaired individuals starting with focus group sessions to ascertain the challenges our target users face in their professional lives. We followed our technology development with a user study to assess the solution and direct our future efforts. We present our findings and conclusions from the study.

espanolLos sistemas de aprendizaje automatico y de IA adquieren cada vez mas capacidades y se utilizan en mas ambitos. Y nosotros, en cuanto disenadores, creadores e investigadores, debemos tener en cuenta tanto las implicaciones positivas como las negativas de su uso. A la luz de lo antedicho, los investigadores del Palo Alto Research Center (PARC) reconocen que no se puede bajar la guardia ante el dano que puede provocar la inteligencia artificial en forma de discriminacion intencionada o inadvertida, trato injusto o lesiones que se pueden infligir a personas o grupos de personas. Puesto que los procesos de decision autonomos y respaldados por IA podrian comportar efectos personales, sociales y medioambientales negativos generalizados, nos proponemos adoptar una postura proactiva de defensa de los derechos humanos, de respeto de la intimidad de los individuos, de proteccion de la informacion personal y de fomento de la libertad de expresion y de la igualdad. La tecnologia no es neutral por naturaleza y refleja las decisiones y las concesiones de los disenadores, los investigadores y los ingenieros al crearla y utilizarla en su trabajo. Los conjuntos de datos a menudo reflejan distorsiones historicas. Las tecnologias de IA que contratan a personas, evaluan su rendimiento laboral, proporcionan atencion sanitaria e imponen multas son ejemplos evidentes de posibles ambitos en los que se pueden producir errores sistematicos en los algoritmos que den lugar a un trato injusto. Casi toda la tecnologia incluye concesiones y encarna los valores y los juicios de las personas que la han creado. Por eso es imprescindible que los investigadores sean conscientes de los juicios de valor que hacen y que sean claros al respecto con todas las partes implicadas. catalaEls sistemes d'aprenentatge automatic i d'IA cada vegada adquireixen mes capacitats i s'utilitzen en mes ambits. I nosaltres, com a dissenyadors, creadors i investigadors, hem de tenir en compte les implicacions del seu us, tant les positives com les negatives. En vista del que acabem d'exposar, els investigadors del Palo Alto Research Center (PARC) reconeixen que no es pot abaixar la guardia davant el dany que la intel·ligencia artificial pot provocar en forma de discriminacio intencionada o inadvertida, tracte injust o lesions que es poden infligir a persones o grups de persones. Com que els processos de decisio autonoms i emparats per IA podrien comportar efectes personals, socials i mediambientals negatius generalitzats, ens proposem adoptar una posicio proactiva de defensa dels drets humans, de respecte de la intimitat dels individus, de proteccio de la informacio personal i de foment de la llibertat d'expressio i de la igualtat. La tecnologia no es neutral per naturalesa i reflecteix les decisions i les concessions dels dissenyadors, els investigadors i els enginyers que la creen i la utilitzen en la seva feina. Els conjunts de dades sovint reflecteixen distorsions historiques. Les tecnologies d'IA que contracten persones, n’avaluen el rendiment laboral, proporcionen atencio sanitaria i imposen multes son exemples evidents de possibles ambits en que es poden produir errors sistematics en els algoritmes que donin lloc a un tracte injust. Gairebe tota la tecnologia inclou concessions i encarna els valors i els judicis de les persones que l'han creat. Per aixo es imprescindible que els investigadors siguin conscients dels judicis de valor que fan i que siguin clars pel que fa al cas amb totes les parts implicades. EnglishAs machine learning and AI systems gain greater capabilities and are deployed more widely, we – as designers, developers, and researchers – must consider both the positive and negative implications of their use. In light of this, PARC’s researchers recognize the need to be vigilant against the potential for harm caused by artificial intelligence through intentional or inadvertent discrimination, unjust treatment, or physical danger that might occur against individuals or groups of people. Because AI-supported and autonomous decision making has the potential for widespread negative personal, social, and environmental effects, we aim to take a proactive stance to uphold human rights, respect individuals’ privacy, protect personal data, and enable freedom of expression and equality. Technology is not inherently neutral and reflects decisions and trade-offs made by the designers, researchers, and engineers developing it and using it in their work. Datasets often reflect historical biases. AI technologies that hire people, evaluate their job performance, deliver their healthcare, and mete out penalties are obvious examples of possible areas for systematic algorithmic errors that result in unfair or unjust treatment. Because nearly all technology includes trade-offs and embodies the values and judgments of the people creating it, it is imperative that researchers are aware of the value judgments they make and are transparent about them with all stakeholders involved.

As machine learning and AI systems gain greater capabilities and are deployed more widely, we – as designers, developers, and researchers – must consider both the positive and negative implications of their use. In light of this, PARC’s researchers recognize the need to be vigilant against the potential for harm caused by artificial intelligence through intentional or inadvertent discrimination, unjust treatment, or physical danger that might occur against individuals or groups of people. Because AI-supported and autonomous decision making has the potential for widespread negative personal, social, and environmental effects, we aim to take a proactive stance to uphold human rights, respect individuals’ privacy, protect personal data, and enable freedom of expression and equality. Technology is not inherently neutral and reflects decisions and trade-offs made by the designers, researchers, and engineers developing it and using it in their work. Datasets often reflect historical biases. AI technologies that hire people, evaluate their job performance, deliver their healthcare, and mete out penalties are obvious examples of possible areas for systematic algorithmic errors that result in unfair or unjust treatment. Because nearly all technology includes trade-offs and embodies the values and judgments of the people creating it, it is imperative that researchers are aware of the value judgments they make and are transparent about them with all stakeholders involved.

Otto is a call center agent designed to help cell phone and mobile provider customers. It has the twin goals of automating call center operations while maintaining a high level of customer satisfaction. It is intended to engage with human users in mixed-initiative dialogs to answer questions, explain procedures, and help with diagnostic troubleshooting. Within the domain of mobile devices, it collaborates with customers on specific tasks but maintains a degree of flexibility and naturalness in the interaction. The development of Otto incorporated aspects of human conversation to improve the quality and likelihood of the success of its interactions.

This poster abstract presents a new word cloud technique, the Fisheye Word Cloud, for exploring time-series data in a focused+context approach to analyzing word data. Our design has two features: cursor-centric layout and word cloud generation on demand. We conducted a validation study to evaluate how our Fisheye Word Cloud influences user performance in comparison tasks of time-series data. Based on task completion time and a TLX-based Likert-style questionnaire, we found the Fisheye Word Cloud has faster task completion time and a better user satisfaction level than the alternative we reviewed.

In a separate study, we were interested in understanding people's Q&A habits on Twitter. Finding questions within Twitter turned out to be a difficult challenge, so we considered applying some traditional NLP approaches to the problem. On the one hand, Twitter is full of idiosyncrasies, which make processing it difficult. On the other, it is very restricted in length and tends to employ simple syntactic constructions, which could help the performance of NLP processing. In order to find out the viability of NLP and Twitter, we built a pipeline of tools to work specifically with Twitter input for the task of finding questions in tweets. This work is still preliminary, but in this paper we discuss the techniques we used and the lessons we learned.

Our work addresses the needs of multiple information workers collaborating on joint projects, which typically require finding, analyzing, and synthesizing information from heterogeneous sources. We report on iterative design, implementation, and assessment of collaborative tools for sensemaking tasks. Our goal is flexible, lightweight tools that both facilitate the activities done individually and lower the costs of effective collaboration. We suggest several approaches to enhance such collaborative sensemaking tools. These approaches include explicit representation of multiple team activities, integrated support for synchronous communication, and views of collected information that are tuned to both the reading and organizing phases of sensemaking. We present an integrated pair of tools, ContextBar and ContextBook, which illustrate these approaches, and describe the results from a formative evaluation of these tools.

Internet applications are being used for more and more important business and personal purposes. Despite efforts to lock down web servers and isolate databases, there is an inherent problem in the web application architecture that leaves databases necessarily exposed to possible attack from the Internet. We propose a new design that removes the web server as a trusted component of the architecture and provides an extra layer of protection against database attacks. We have created a prototype system that demonstrates the feasibility of the new design.

Postfix is a Mail Transfer Agent (MTA): software that mail servers use to route email. Postfix is highly respected by experts for its secure design and tremendous reliability. And new users like it because it's so simple to configure. In fact, Postfix has been adopted as the default MTA on Mac OS X. It is also compatible with sendmail, so that existing scripts and programs continue to work seamlessly after it is installed. Postfix was written by well-known security expert Wietse Venema, who reviewed this book intensively during its entire development. Author Kyle Dent covers a wide range of Postfix tasks, from virtual hosting to controls for unsolicited commercial email. While basic configuration of Postfix is easy, every site has unique needs that call for a certain amount of study. This book, with careful background explanations and generous examples, eases readers from the basic configuration to the full power of Postfix. It discusses the Postfix interfaces to various tools that round out a fully scalable and highly secure email system. These tools include POP, IMAP, LDAP, MySQL, Simple Authentication and Security Layer (SASL), and Transport Layer Security (TLS, an upgrade of SSL). A reference section for Postfix configuration parameters and an installation guide are included. Topics include: Basic installation and configurationDNS configuration for emailWorking with POP/IMAP serversHosting multiple domains (virtual hosting)Mailing listsHandling unsolicited email (spam blocking)Security through SASL and TLS From compiling and installing Postfix to troubleshooting, Postfix: The Definitive Guide offers system administrators and anyone who deals with Postfix an all-in-one, comprehensive tutorial and reference to this MTA.